Statistical Methods for Multi-modal Neuroimaging Data: Techniques for the Combined Analysis of Brain Function and Structure 公开

Abstract

Recent innovations in neuroimaging technology have provided opportunities for researchers to investigate human brain function and structure, improving our understanding of psychiatric disorders, clinical diagnosis, and neural networks. Brain imaging data have massive dimensionality and are marked by complex spatial and temporal correlations, which pose challenges for statistical modeling. Our objective is to develop novel statistical methods for high-dimensional neuroimaging data to explore the underlying complex neural processing in the human brain. Specifically, we propose three new statistical frameworks: (i) to determine the functional coherence and the associated hierarchical relationships between brain regions using combined functional and structural data; (ii) to robustly characterize co-activation patterns and functional networks using a novel meta-analytic approach; and (iii) to predict the disease status using imaging data from different modalities.

Rapid development in neuroimaging allows researchers to study the connectivity in the human brain by examining the anatomical circuitry and functional relationships between brain regions. We present a unified Bayesian framework for analyzing functional connectivity utilizing the knowledge of associated structural connections, which extends an approach by Patel et al. (2006a) that considers only functional data. We demonstrate the use of our Bayesian model using fMRI and DTI data from a study of auditory processing.

Meta analysis plays an important role in neuroimaging research. Several approaches have been developed to determine the consistency in activated brain regions for imaging studies. We focus on identifying the functional co-activation patterns and building a non-directed functional network in the human brain. We adopt a penalized likelihood approach to impose sparsity on the covariance matrix for region-level peak activations, which is used to construct a brain network. We apply our proposed method to a meta analysis of 162 functional neuroimaging studies on emotions.

Relating disease status to imaging data increases the clinical significance of neuroimaging studies. We propose a Bayesian hierarchical model to predict the disease status using both the functional and structural imaging scans. We consider a two-level brain parcellation, and take into account the correlations between voxels from different levels. We conduct both whole-brain and voxel-level prediction, and apply our model to a study of Parkinson's disease.

Table of Contents

Contents

1 Introduction 1

1.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

1.2 An Introduction to the Human Brain . . . . . . . . . . . . . . . . . . 2

1.3 Functional Neuroimaging . . . . . . . . . . . . . . . . . . . . . . . . . 5

1.3.1 Functional Magnetic Resonance Imaging (fMRI) . . . . . . . . 6

1.3.2 Preprocessing . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

1.3.3 Analysis of Functional Connectivity . . . . . . . . . . . . . . . 10

1.4 Structural Neuroimaging . . . . . . . . . . . . . . . . . . . . . . . . . 14

1.4.1 Diffusion Tensor Imaging (DTI) . . . . . . . . . . . . . . . . . 14

1.4.2 Preprocessing . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

1.4.3 Analysis of Structural Connectivity . . . . . . . . . . . . . . . 15

1.5 Prediction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

1.6 Motivating Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

1.6.1 An fMRI Study on Auditory Spatial-Cueing Task . . . . . . . 17

1.6.2 A Meta Analysis of Emotions . . . . . . . . . . . . . . . . . . 18

1.6.3 Parkinson's Disease Data . . . . . . . . . . . . . . . . . . . . . 19

1.7 Proposed Research . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20

1.7.1 Modeling Functional Connectivity with Incorporation of Structural

Connectivity . . . . . . . . . . . . . . . . . . . . . . . . 20

1.7.2 A Graphical Model for Count Data: A Meta Analysis of Functional

Co-activation Patterns in Imaging Studies . . . . . . . . 21

1.7.3 A Bayesian Spatial Model to Predict Disease Status Using Imaging

Data from Different Modalities . . . . . . . . . . . . . . . 21

2 Modeling Functional Connectivity in the Human Brain with Incorporation

of Structural Connectivity 23

2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

2.2 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

2.2.1 Experimental Data . . . . . . . . . . . . . . . . . . . . . . . . 26

2.2.2 Image Acquisition and Data Preprocessing . . . . . . . . . . . 27

2.2.3 Determining Regional Activity . . . . . . . . . . . . . . . . . . 27

2.2.4 Determining Structural Connectivity . . . . . . . . . . . . . . 28

2.3 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29

2.3.1 Joint Activation and Structural Connectivity . . . . . . . . . . 30

2.3.2 Functional Coherence and Ascendancy . . . . . . . . . . . . . 31

2.3.3 Bayesian Model . . . . . . . . . . . . . . . . . . . . . . . . . . 33

2.3.4 Posterior Sampling . . . . . . . . . . . . . . . . . . . . . . . . 35

2.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37

2.4.1 Auditory Data Results . . . . . . . . . . . . . . . . . . . . . . 37

2.4.2 Simulation Results . . . . . . . . . . . . . . . . . . . . . . . . 40

2.5 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43

3 Identifying Functional Co-activation Patterns in Neuroimaging Studies

via Poisson Graphical Models 48

3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48

3.2 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51

3.3 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52

3.3.1 The Bivariate Model . . . . . . . . . . . . . . . . . . . . . . . 52

3.3.2 The Multivariate Model . . . . . . . . . . . . . . . . . . . . . 55

3.3.3 Tuning Parameter . . . . . . . . . . . . . . . . . . . . . . . . . 57

3.3.4 Statistical Testing . . . . . . . . . . . . . . . . . . . . . . . . . 59

3.3.5 Graph Theoretical Properties of the Network . . . . . . . . . . 60

3.3.6 Initial Values of EM Algorithm . . . . . . . . . . . . . . . . . 60

3.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61

3.4.1 A Meta Analysis of Functional Neuroimaging Studies . . . . . 61

3.4.2 Simulation Studies . . . . . . . . . . . . . . . . . . . . . . . . 66

3.5 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70

4 A Bayesian Spatial Model to Predict Disease Status Using Imaging

Data from Various Modalities 73

4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73

4.2 Parkinson's Disease Data . . . . . . . . . . . . . . . . . . . . . . . . . 75

4.3 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76

4.3.1 Model and Estimation . . . . . . . . . . . . . . . . . . . . . . 76

4.3.2 Prediction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80

4.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86

4.4.1 Parkinson's Disease Data . . . . . . . . . . . . . . . . . . . . . 86

4.4.2 Simulation Studies . . . . . . . . . . . . . . . . . . . . . . . . 89

4.5 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90

5 Summary and Future Work 93

5.1 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93

5.2 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94

Appendices 96

Bibliography 108

About this Dissertation

Rights statement

• Permission granted by the author to include this thesis or dissertation in this repository. All rights reserved by the author. Please contact the author for information regarding the reproduction and use of this thesis or dissertation.

School	Laney Graduate School
Department	Biostatistics
Degree	PhD
Submission	Dissertation
Language	English
Research Field	Biology, Biostatistics Statistics Biology, Neuroscience
关键词	Structural Connectivity Neuroimaging Spatial Correlation Functional Co-activation Functional Connectivity Meta Analysis Bayesian Modeling Prediction
Committee Chair / Thesis Advisor	Bowman, F Dubois, Emory University
Committee Members	Rilling, James K, Emory University Waller, Lance, Emory University Kang, Jian , Emory University

最新修改

Primary PDF

Thumbnail	Title	Date Uploaded	Actions
	Statistical Methods for Multi-modal Neuroimaging Data: Techniques for the Combined Analysis of Brain Function and Structure ()	2018-08-28 12:45:37 -0400	Press to Select an action Download

Supplemental Files

Thumbnail	Title	Date Uploaded	Actions